<!DOCTYPE HTML>
<html>
<head>
<meta charset="UTF-8">
<title>Examples 5.6,5.8: An l_p norm approximation problem</title>
<link rel="canonical" href="/Users/mcgrant/Repos/CVX/examples/cvxbook/Ch05_duality/html/norm_approx.html">
<link rel="stylesheet" href="../../../examples.css" type="text/css">
</head>
<body>
<div id="header">
<h1>Examples 5.6,5.8: An l_p norm approximation problem</h1>
Jump to:&nbsp;&nbsp;&nbsp;&nbsp;
<a href="#source">Source code</a>&nbsp;&nbsp;&nbsp;&nbsp;
<a href="#output">Text output</a>
&nbsp;&nbsp;&nbsp;&nbsp;
Plots
&nbsp;&nbsp;&nbsp;&nbsp;<a href="../../../index.html">Library index</a>
</div>
<div id="content">
<a id="source"></a>
<pre class="codeinput">
<span class="comment">% Boyd &amp; Vandenberghe "Convex Optimization"</span>
<span class="comment">% Jo&euml;lle Skaf - 08/23/05</span>
<span class="comment">%</span>
<span class="comment">% The goal is to show the following problem formulations give all the same</span>
<span class="comment">% optimal residual norm ||Ax - b||:</span>
<span class="comment">% 1)        minimize    ||Ax - b||</span>
<span class="comment">% 2)        minimize    ||y||</span>
<span class="comment">%               s.t.    Ax - b = y</span>
<span class="comment">% 3)        maximize    b'v</span>
<span class="comment">%               s.t.    ||v||* &lt;= 1  , A'v = 0</span>
<span class="comment">% 4)        minimize    1/2 ||y||^2</span>
<span class="comment">%               s.t.    Ax - b = y</span>
<span class="comment">% 5)        maximize    -1/2||v||*^2 + b'v</span>
<span class="comment">%               s.t.    A'v = 0</span>
<span class="comment">% where ||.||* denotes the dual norm of ||.||</span>

<span class="comment">% Input data</span>
randn(<span class="string">'state'</span>,0);
n = 4;
m = 2*n;
A = randn(m,n);
b = randn(m,1);
p = 2;
q = p/(p-1);

<span class="comment">% Original problem</span>
fprintf(1,<span class="string">'Computing the optimal solution of problem 1... '</span>);

cvx_begin <span class="string">quiet</span>
    variable <span class="string">x(n)</span>
    minimize ( norm ( A*x - b , p) )
cvx_end

fprintf(1,<span class="string">'Done! \n'</span>);
opt1 = cvx_optval;

<span class="comment">% Reformulation 1</span>
fprintf(1,<span class="string">'Computing the optimal solution of problem 2... '</span>);

cvx_begin <span class="string">quiet</span>
    variables <span class="string">x(n)</span> <span class="string">y(m)</span>
    minimize ( norm ( y , p ) )
    A*x - b == y;
cvx_end

fprintf(1,<span class="string">'Done! \n'</span>);
opt2 = cvx_optval;

<span class="comment">% Dual of reformulation 1</span>
fprintf(1,<span class="string">'Computing the optimal solution of problem 3... '</span>);

cvx_begin <span class="string">quiet</span>
    variable <span class="string">nu(m)</span>
    maximize ( b'*nu )
    norm( nu , q ) &lt;= 1;
    A'*nu == 0;
cvx_end

fprintf(1,<span class="string">'Done! \n'</span>);
opt3 = cvx_optval;

<span class="comment">% Reformulation 2</span>
fprintf(1,<span class="string">'Computing the optimal solution of problem 4... '</span>);

cvx_begin <span class="string">quiet</span>
    variables <span class="string">x(n)</span> <span class="string">y(m)</span>
    minimize ( 0.5 * square_pos ( norm ( y , p ) ) )
    A*x - b == y;
cvx_end

fprintf(1,<span class="string">'Done! \n'</span>);
opt4 = (2*cvx_optval).^(.5);

<span class="comment">% Dual of reformulation 2</span>
fprintf(1,<span class="string">'Computing the optimal solution of problem 5... '</span>);

cvx_begin <span class="string">quiet</span>
    variable <span class="string">nu(m)</span>
    maximize ( -0.5 * square_pos ( norm ( nu , q ) ) + b'*nu )
    A'*nu == 0;
cvx_end

fprintf(1,<span class="string">'Done! \n'</span>);
opt5 = (2*cvx_optval).^(0.5);

<span class="comment">% Displaying results</span>
disp(<span class="string">'------------------------------------------------------------------------'</span>);
disp(<span class="string">'The optimal residual values for problems 1,2,3,4 and 5 are respectively:'</span>);
[ opt1 opt2 opt3 opt4 opt5 ]'
disp(<span class="string">'They are equal as expected!'</span>);
</pre>
<a id="output"></a>
<pre class="codeoutput">
Computing the optimal solution of problem 1... Done! 
Computing the optimal solution of problem 2... Done! 
Computing the optimal solution of problem 3... Done! 
Computing the optimal solution of problem 4... Done! 
Computing the optimal solution of problem 5... Done! 
------------------------------------------------------------------------
The optimal residual values for problems 1,2,3,4 and 5 are respectively:

ans =

    1.8371
    1.8371
    1.8371
    1.8371
    1.8371

They are equal as expected!
</pre>
</div>
</body>
</html>